Amazing Science

ChatGPT recently passed all three parts of the U.S. Medical Licensing Examination.

ChatGPT, the generative AI juggernaut, is getting a lot smarter when it comes to health care. A lot of clinical diagnoses and decisions could someday be made by machines, rather than by human doctors. As the researchers note, second-year medical students often spend hundreds of hours preparing for Part 1, while Part 3 usually is taken by medical school graduates.

The big surprise was that ChatGPT could perform so well without ever having been trained on a medical dataset.

• One caveat is that researchers excluded a set of "indeterminate" answers, as it appears that ChatGPT was programmed to avoid providing what could be construed as medical advice.
• "Those answers were so general that it was hard to say if they were right or wrong," explains paper co-author Morgan Cheatham, an investor with Bessemer Venture Partners and current medical school student at Brown University.

Generative AI remains in the early innings, so for now it'll augment medical work rather than replace it.

• Ansible, for example, is using ChatGPT to help explain certain concepts to patients, after review by a trained professional.

What's next: Over time, perhaps it could be applied to wellness checks and other general practitioner tasks.

• Once the technology moves beyond just text, it could incorporate data inputs like vocal tone, body language and facial expressions.
• One benefit would be the immediate incorporation of a patient's medical records. Sometimes physicians only have moments to scan a lifetime of charts.

Reality check: Don't expect a machine to autonomously diagnose patients anytime soon. AI models like ChatGPT sometimes make confident assertions that turn out to be false — which could prove dangerous in medical applications. "I think we're in the middle of a 20-year arc, kind of like what we already saw with finance," says Vijay Pande, a health care investor with Andreessen Horowitz and adjunct professor of bioengineering at Stanford University.

• "In 2000, it was insane to think that a computer could beat a master trader on Wall Street. Today, it's insane to think that master trader could beat a computer."

The bottom line: Plenty of people rely on "Dr. Google" for their medical information needs. In the future, they may turn to "Dr. ChatGPT."

Combining Wolfram Alpha and GPT-3 could potentially create a powerful tool that could answer complex questions and generate human-like text. It could be used to provide in-depth explanations and discussions on a wide range of topics, and could be used to engage in natural-sounding conversations.

Accessing Wolfram|Alpha's computational knowledge with ChatGPT--an ideal combination of precise computation with human-like expression of ideas. Stephen Wolfram explains how.

It’s always amazing when things suddenly “just work”. It happened to us with Wolfram|Alpha back in 2009. It happened with our Physics Project in 2020. And it’s happening now with OpenAI’s ChatGPT. Stephen Wolfram has been tracking neural net technology for about 43 years and he finds the performance of ChatGPT thoroughly remarkable. Suddenly, there is a system that can successfully generate text about almost anything—that’s very comparable to what humans might write. It’s impressive, and useful and its success is probably going to tell us some very fundamental things about the nature of human thinking.

But while ChatGPT is a remarkable achievement in automating the doing of major human-like things, not everything that’s useful to do is quite so “human like”. Some of it is instead more formal and structured. And indeed one of the great achievements of our civilization over the past several centuries has been to build up the paradigms of mathematics, the exact sciences—and, most importantly, now computation—and to create a tower of capabilities quite different from what pure human-like thinking can achieve.

Stephen Wolfram himself has been deeply involved with the computational paradigm for many decades, in the singular pursuit of building a computational language to represent as many things in the world as possible in formal symbolic ways. His goal has always been to build a system that can “computationally assist”—and augment—what I and others want to do. He thinks about things as a human but can also immediately use Wolfram Language and Wolfram|Alpha to tap into a kind of unique “computational superpower” that lets him perform all sorts of beyond-human things. It’s a tremendously powerful way of working and not just important for humans but equally, if not more, important for human-like AIs as well—immediately giving them computational knowledge superpowers, that leverage the non-human-like power of structured computation and structured knowledge.

We’ve just started exploring what this means for ChatGPT. But it’s pretty clear that wonderful things are now possible. Wolfram|Alpha does something very different from ChatGPT, in a very different way. But they have a common interface: natural language. And this means that ChatGPT can “talk to” Wolfram|Alpha just like humans do—with Wolfram|Alpha turning the natural language it gets from ChatGPT into precise, symbolic computational language on which it can apply its computational knowledge power.

For decades there’s been a dichotomy in thinking about AI between “statistical approaches” of the kind ChatGPT uses, and “symbolic approaches” that are in effect the starting point for Wolfram|Alpha. But now—thanks to the success of ChatGPT—as well as all the work we’ve done in making Wolfram|Alpha understand natural language—there’s finally the opportunity to combine these to make something much stronger than either could ever achieve on their own.

More than 90% of the world's population is projected to face increased risks from the compound impacts of extreme heat and drought, potentially widening social inequalities as well as undermining the natural world's ability to reduce CO2 emissions in the atmosphere—according to a study from Oxford's School of Geography.

Global warming is projected to intensify these hazards ten-fold globally under the highest emission pathway, says the report, published in Nature Sustainability. In the wake of record temperatures in 2022, from London to Shanghai, continuing rising temperatures are projected around the world. When assessed together, the linked threats of heat and drought represent a significantly higher risk to society and ecosystems than when either threat is considered independently, according to the paper by Dr. Jiabo Yin, a visiting researcher from Wuhan University and Oxford Professor Louise Slater.

These joint threats may have severe socio-economic and ecological impacts which could aggravate socio inequalities, as they are projected to have more severe impacts on poorer people and rural areas. According to the research, "The frequency of extreme compounding hazards is projected to intensify tenfold globally due to the combined effects of warming and decreases in terrestrial water storage, under the highest emission scenario.

Over 90% of the world population and GDP is projected to be exposed to increasing compounding risks in the future climate, even under the lowest emission scenario." Dr. Yin says, "By using simulations from a large model…and a new machine-learning generated carbon budget dataset, we quantify the response of ecosystem productivity to heat and water stressors at the global scale." He maintains this shows the devastating impact of the compound threat on the natural world—and international economies. He says, limited water availability will hit the ability of "carbon sinks"—natural biodiverse regions—to take in carbon emissions and emit oxygen.

Professor Slater says, "Understanding compounding hazards in a warming Earth is essential for the implementation of the UN Sustainable Development Goals (SDGs), in particular SDG13 that aims to combat climate change and its impacts. By combining atmospheric dynamics and hydrology, we explore the role of water and energy budgets in causing these extremes."

A team from Nagoya University in Japan has observed, for the first time, the energy transferring from resonant electrons to whistler-mode waves in space. Their findings offer direct evidence of previously theorized efficient growth, as predicted by the non-linear growth theory of waves. This should improve our understanding of not only space plasma physics but also space weather, a phenomenon that affects satellites. 

When people imagine outer space, they often envision it as a perfect vacuum. In fact, this impression is wrong because the vacuum is filled with charged particles. In the depths of space, the density of charged particles becomes so low that they rarely collide with each other. Instead of collisions, the forces related to the electric and magnetic fields filling space, control the motion of charged particles. This lack of collisions occurs throughout space, except for very near to celestial objects, such as stars, moons, or planets. In these cases, the charged particles are no longer traveling through the vacuum of space but instead through a medium where they can strike other particles. 

Around the Earth, these charged-particle interactions generate waves, including electromagnetic whistler-mode waves, which scatter and accelerate some of the charged particles. When diffuse auroras appear around the poles of planets, observers are seeing the results of an interaction between waves and electrons. Since electromagnetic fields are so important in space weather, studying these interactions should help scientists predict variations in the intensity of highly energetic particles. This might help protect astronauts and satellites from the most severe effects of space weather.  

A team comprising Assistant Professor Naritoshi Kitamura and Professor Yoshizumi Miyoshi of the Institute for Space and Earth Science (ISEE) at Nagoya University, together with researchers from the University of Tokyo, Kyoto University, Tohoku University, Osaka University, and Japan Aerospace Exploration Agency (JAXA), and several international collaborators, mainly used data obtained using low-energy electron spectrometers, called Fast Plasma Investigation-Dual Electron Spectrometers, on board NASA’s Magnetospheric Multiscale spacecraft. They analyzed interactions between electrons and whistler-mode waves, which were also measured by the spacecraft. By applying a method of using a wave particle interaction analyzer, they succeeded in directly detecting the ongoing energy transfer from resonant electrons to whistler-mode waves at the location of the spacecraft in space. From this, they derived the growth rate of the wave. The researchers published their results in Nature Communications. 

The Standard Model of particle physics is often visualized as a table, similar to the periodic table of elements, and used to describe particle properties, such as mass, charge and spin. The table is also organized to represent how these teeny, tiny bits of matter interact with the fundamental forces of nature.

But it didn’t begin as a table. The grand theory of almost everything actually represents a collection of several mathematical models that proved to be timeless interpretations of the laws of physics.

Here is a brief tour of the topics covered in this gargantuan equation.

The whole thing

This version of the Standard Model is written in the Lagrangian form. The Lagrangian is a fancy way of writing an equation to determine the state of a changing system and explain the maximum possible energy the system can maintain.  Technically, the Standard Model can be written in several different formulations, but, despite appearances, the Lagrangian is one of the easiest and most compact ways of presenting the theory.

Created by AI

https://twitter.com/debarghya_das/status/1610470866713972737

Smart summary:
The thread discusses the upcoming "Prompt Palettes" feature for ChatGPT, which will provide users with pre-written text prompts to help with tasks like formatting raw text, summarizing text, and serving as a programming assistant. The feature will use OpenAI's GPT-3 Codex models, and is similar to a "brushes" feature being developed by Github Copilot Labs.
-------

ChatGPT will soon drop a new feature – Prompt Palettes!

They're pre-written text prompts to perform tasks like
- format raw text to markdown
- summarize text
- be a programming assistant
- add text from a link as context

How it works and EXACT prompts are pre-written bits of text that accentuate a user's input for a specific task.

It's like a magic button for text explained only by pre-written instructions. Here are the prompts to "format" and "summarize". You should be able to add your own too.

— ChatGPT Coding Assistant —

is slightly different from Prompt Palettes but seems like a pre-written addendum to a query to serve as a one-shot way to focus on a specific vertical task.

This might be forcing the use of GPT-3: Codex explicitly– https://t.co/GddMvqG3pU

Prompt palettes act on a specific message, and ChatGPT is adding a nifty "Add text from link" feature which will allow you to, say, summarize websites easily.

Prompt palettes aren't new! Github Copilot Labs has been working on a similar magic "brushes" feature that integrates directly into VS Code. They use OpenAI's GPT-3 Codex models too.

https://t.co/8gOp9L17Bx

Prompt Palettes will bring these powerful new LLM features to the mainstream of 1 million users! 2023 has just begun for AI.

https://t.co/04GwCmOQr8

Thanks to @eeeziii for the idea of reverse-engineering ChatGPT (he did this too)!

ChatGPT is a minified React app with chunked JS that uses Server Sent Events with /conversations to stream the meat of the output. It uses text-davinci-002-render with 4097 max tokens.

As ground-based and space telescopes improve, astronomers need a color-coded guide to compare Earth’s biological microbes to cold, distant exoplanets to grasp their composition.

As our knowledge of the resilience of life in extreme conditions on Earth grows, it opens up the possibility for new places to search for life, even far beyond our Solar System. Cataloging the biomarkers of organisms from unique and extreme environments on Earth will hopefully enhance the ability to identify signs of life on other worlds.

The new work focuses on the identification of ice-dwelling bacteria and algae on cold exoplanets and exomoons via their color spectra. The work is measuring the pigment spectral signatures of 80 different Earth organisms to build a reference catalog for future characterizations of directly observed worlds beyond our Solar System. The depicted hypothetical icy Earth-like exoplanet is covered in the types of strains of cold-loving bacteria that occupy Earth's frozen landscapes. Though we do not expect the spectra of terrestrial psychrophilic strains to provide a perfect spectral match with such hypothetical organisms on other worlds, similarities in the environmental conditions of icy exoplanets—low temperatures, low concentrations of nutrients and carbon sources, enhanced dryness, increased radiation—may have led to a common set of environmentally driven physiological adaptations and pigments of carbon-based life that could be recognized spectrally throughout the Universe.

Laser powder bed fusion, a 3D-printing technique, offers potential in the manufacturing industry, particularly when fabricating nickel-titanium shape memory alloys with complex geometries. Although this manufacturing technique is attractive for applications in the biomedical and aerospace fields, it has rarely showcased the superelasticity required for specific applications using nickel-titanium shape memory alloys. Defects generated and changes imposed onto the material during the 3D-printing process prevented the superelasticity from appearing in 3D-printed nickel-titanium.

Researchers from Texas A&M University recently showcased superior tensile superelasticity by fabricating a shape memory alloy through laser powder bed fusion, nearly doubling the maximum superelasticity reported in literature for 3D printing.

This study was recently published in vol. 229 of the Acta Materialia journal.

Nickel-titanium shape memory alloys have various applications due to their ability to return to their original shape upon heating or upon removal of the applied stress. Therefore, they can be used in biomedical and aerospace fields for stents, implants, surgical devices and aircraft wings. However, developing and properly fabricating these materials requires extensive research to characterize functional properties and examine the microstructure.

"Shape memory alloys are smart materials that can remember their high-temperature shapes," said Dr. Lei Xue, a former doctoral student in the Department of Materials Science and Engineering and the first author of the publication. "Although they can be utilized in many ways, fabricating shape memory alloys into complex shapes requires fine-tuning to ensure the material exhibits the desired properties."

GPT-4 is the fourth version of the GPT, one of the most advanced and anticipated AI models. GPT-3 had a significant boost in performance compared to GPT-2.

Since 2018 when GPT-1 was released, OpenAI has followed the “the bigger, the better” strategy. GPT-1 had 117 million parameters, GPT-2 had 1.2 billion parameters, and GPT-3 raised the number even further to 175 billion parameters. It means that the GPT-3 model has 100 times more parameters than GPT-2. GPT-3 is a very large model size, with 175 billion parameters.

In an August 2021 interview with Wired, Andrew Feldman, founder and CEO of Cerebras, a company that partners with OpenAI to train the GPT model, mentioned that GPT-4 will be about 100 trillion parameters. It might seem like GPT-4 will be 100x more powerful than GPT-3. 100 trillion parameters is a low estimate for the number of neural connections in the human brain. If GPT-4 will have 100 trillion parameters, it will match the human brain in terms of parameters. No surprise that this made people so excited.

In reality, model size is not directly related to the quality of the outcome it produces. The number of parameters does not necessarily correlate with the performance that an AI model has. It’s only one factor that impacts the model performance. Right now, we have much larger AI models than GPT-3, but they aren’t the best in terms of performance. For example, Megatron-Turing NLG, built by Nvidia and Microsoft, has more than 500B parameters, and it’s the largest model right now. But despite that MT-NLG isn’t the best in terms of performance. The smaller model can reach higher performance levels. Also, the larger the model, the more expensive to fine-tune it. GPT3 was hard enough and expensive to train, but if you increase the model’s size by 100x, it will be extremely expensive in terms of computational power and the amount of training data required for the model.

There Is a low chance that OpenAI will have 100T parameters in GPT-4 because just increasing the number of training parameters won’t cause any drastic improvement if training data is also not increased proportionally. Large models are often unoptimized, take Megatron-Turing NLG as an example. It is very expensive to train the model, and often companies have to make a trade between AI model accuracy and the cost of training. For example, GPT-3 was trained only once, and, despite the errors in the AI model, OpenAI wasn’t able to train the model again due to unaffordable costs. It all means that OpenAI will likely start to avoid the “bigger is better” approach and focus on the quality of the model itself. Most likely, GPT-4 would be roughly GPT -3’s size.

What is more interesting is that OpenAI will likely shift the focus toward other aspects that impact the model performance, such as algorithms and alignment. GPT-4 might be the first large AI model with sparsity at its core. Sparse models use conditional computation to reduce computing costs — not all neurons in the AI model are active at any given time. The model can easily scale beyond a trillion parameters without incurring high computing costs. Sparse models also better understand the context — they can keep a lot more choices of “next word/sentence” based on what the user provided. As a result, sparse models are more similar to actual human thinking than their predecessors.

The idea started as an essay 20 years ago. This essay was about the principle of photonic propulsion, the idea that it is possible to move space vehicles thanks to sunlight. When a photon reflects on a surface, it transfers a small momentum, exerting pressure on it. Deploying a huge mirror in space would would be all you need to use this photonic pressure to move a vehicle.

This is the principle of the solar sail and the subject of the Gama project which started in August 2020. A solar sail could theoretically accelerate to speeds never before achieved by objects created by humanity. This efficient and economical propulsion method could offer new opportunities in terms of space missions.

And today is an exceptional day for this project and for the whole team. The satellite which is the size of 5 milk bottles (6U) and weighs 12 kg is going to be put on a low earth orbit at 550 km (LEO). It embeds both a "Bus" part containing all the system elements (energy, communication) developed by

NanoAvionics and a "Sail" part on which we have concentrated our efforts. The sail is made of a reflective material (aluminized polyimide: CP1) extremely thin (2.5 microns). It is folded origami-style in a winder and is equipped with 4 Tungsten weights fixed at its ends.

When the satellite leaves the Falcon 9 rocket, a first test phase will begin, followed by a deployment phase. A slow rotation of several days will allow to unfold the 73.3 m2 of the sail. It is the centrifugal force which will make it possible to unroll and to rigidify the sail.

This first mission will simply allow the scientific team to test the deployment of the sail. At this altitude, there is still too much air slowing down the sail and making impossible to navigate it. They will have to wait for our next mission "Gama Beta" in 2024 and the deployment of the sail in high orbit to be able to test the navigation to make the satellite move thanks to the photonic pressure. From that moment on, the scientists will judge whether the technology is mature enough for its commercialization for deep space exploration and exploitation missions.

Prompt-based learning has emerged as a successful paradigm in natural language processing, where a single general-purpose language model can be instructed to perform any task specified by input prompts. Yet task specification in robotics comes in various forms, such as imitating one-shot demonstrations, following language instructions, and reaching visual goals. They are often considered different tasks and tackled by specialized models.

This present work shows that a wide spectrum of robot manipulation tasks can be expressed with multimodal prompts, interleaving textual and visual tokens. The robotic team designed a transformer-based generalist robot agent, VIMA, that processes these prompts and outputs motor actions auto-regressively. To train and evaluate VIMA, they developed a new simulation benchmark with thousands of procedurally-generated tabletop tasks with multimodal prompts, 600K+ expert trajectories for imitation learning, and four levels of evaluation protocol for systematic generalization. VIMA was able to achieve strong scalability in both model capacity and data size. It outperformed prior SOTA methods in the hardest zero-shot generalization setting by up to 2.9 times task success rate given the same training data. With 10 times less training data, VIMA still performed 2.7 times better than the top competing approach.

Cat genes reveal how invention of agriculture bonded cats with people in ancient Mesopotamia, leading to worldwide feline migration with humans.

Nearly 10,000 years ago, humans settling in the Fertile Crescent, the areas of the Middle East surrounding the Tigris and Euphrates rivers, made the first switch from hunter-gatherers to farmers. They developed close bonds with the rodent-eating cats that conveniently served as ancient pest-control in society's first civilizations.

A new study at the University of Missouri found this lifestyle transition for humans was the catalyst that sparked the world's first domestication of cats, and as humans began to travel the world, they brought their new feline friends along with them. Leslie A. Lyons, a feline geneticist and Gilbreath-McLorn endowed professor of comparative medicine in the MU College of Veterinary Medicine, collected and analyzed DNA from cats in and around the Fertile Crescent area, as well as throughout Europe, Asia and Africa, comparing nearly 200 different genetic markers. "One of the DNA main markers we studied were microsatellites, which mutate very quickly and give us clues about recent cat populations and breed developments over the past few hundred years," Lyons said. "Another key DNA marker we examined were single nucleotide polymorphisms, which are single-based changes all throughout the genome that give us clues about their ancient history several thousands of years ago. By studying and comparing both markers, we can start to piece together the evolutionary story of cats."

Lyons added that while horses and cattle have seen various domestication events caused by humans in different parts of the world at various times, her analysis of feline genetics in the study strongly supports the theory that cats were likely first domesticated only in the Fertile Crescent before migrating with humans all over the world. After feline genes are passed down to kittens throughout generations, the genetic makeup of cats in western Europe, for example, is now far different from cats in southeast Asia, a process known as 'isolation by distance.'

"We can actually refer to cats as semi-domesticated, because if we turned them loose into the wild, they would likely still hunt vermin and be able to survive and mate on their own due to their natural behaviors," Lyons said. "Unlike dogs and other domesticated animals, we haven't really changed the behaviors of cats that much during the domestication process, so cats once again prove to be a special animal."

Lyons, who has researched feline genetics for more than 30 years, said studies like this also support her broader research goal of using cats as a biomedical model to study genetic diseases that impact both cats and people, such as polycystic kidney disease, blindness and dwarfism.

"Comparative genetics and precision medicine play key roles in the 'One Health' concept, which means anything we can do to study the causes of genetic diseases in cats or how to treat their ailments can be useful for one day treating humans with the same diseases," Lyons said. "I am building genetic tools, genetic resources that ultimately help improve cat health. When building these tools, it is important to get a representative sample and understand the genetic diversity of cats worldwide so that our genetic toolbox can be useful to help cats all over the globe, not just in one specific region."

CureVac, the biggest defender of ‘unmodified’ mRNA for vaccines against infectious disease has had a change of heart.

It is only a small chemical tweak, but it has triggered huge division in the mRNA research community. Proponents argue that a simple chemical modification to the backbone of mRNA is crucial to the success of mRNA vaccines, allowing them to trigger a potent immune response without massive side effects. The blockbuster COVID-19 vaccines made by Pfizer–BioNTech and Moderna, both of which have this modification. But some mRNA manufacturers have long held that unmodified mRNA elicits a superior immune response to pathogens — a rift with implications for mRNA vaccine development.

Now the biggest champion of the unmodified approach has changed its tune. CureVac — which is one of the world’s oldest mRNA specialty companies and is based in Tübingen, Germany — is ditching unmodified mRNA, and embracing the modified version for its entire infectious-disease vaccine portfolio after disappointing results from COVID-19 vaccine trials. Modified mRNA is the “best-performing technology for prophylactic vaccines”, Franz-Werner Haas, CureVac’s outgoing chief executive, said in a conference call last week to announce the company’s strategic shift — a retreat from a position it has held since its founding in 2000.

The company’s about-face should help put to rest the decades-long debate among scientists over which type of mRNA is best to protect against influenza, SARS-CoV-2 and other viral threats. CureVac’s pivot in the respiratory-disease arena leaves only a few late-stage products reliant on unmodified mRNA, all of which are COVID-19 vaccine candidates under development in Asia.

Vaccines designed around mRNA work by giving the body instructions on how to make viral proteins. Those proteins then train the immune system to produce antibodies that can disarm invading pathogens. But the process only works if the immune system doesn’t see the vaccine mRNA itself as a threat.

That’s where modifications come in. In the mid-2000s, scientists discovered that swapping one of RNA’s four basic building blocks, called nucleotides, for a related analogue helped to make the mRNA less conspicuous to the immune system. Vaccines could then be given at higher doses, allowing for maximal antibody responses with fewer of the icky side effects — fevers, chills, aches — that result from immune reactions to the mRNA itself.

Fast-evolving COVID variants complicate vaccine updates

For decades, CureVac thought that it could achieve the same effect by fine-tuning the mRNA sequence to avoid tripping the immune alarm. The company claimed that its ‘optimized’ mRNA could elicit the full breadth of desired immune responses without any major safety issues1. But in late-stage testing, CureVac’s experimental jab didn’t yield enough antibodies to thwart emerging variants of SARS-CoV-22. The general consensus: CureVac’s dose was too low but couldn’t be raised because of tolerability concerns.

The company created a next-generation jab that, although still built around unmodified mRNA, seemed to generate more robust immune responses at the same dose used for its earlier vaccine3. The product entered clinical trials in March 2022. But around the same time, a shift in mindset was taking hold at the company.

As Haas told Nature in an interview that month, “There should not be a kind of religious way to see” the issue of mRNA modifications. CureVac and its co-development partner, GSK of London, decided to give modified mRNA a real try, with the idea being, as Haas put it, to “let the data talk”.

In August, 2022, another COVID-19 vaccine candidate from CureVac entered clinical trials, this one updated to reflect the circulating Omicron variant of SARS-CoV-2 — and now with modified mRNA.

Chemical conversion

On 6 January, 2023, CureVac reported preliminary-trial data comparing the two next-gen jabs. At the same dose, a shot of the modified mRNA elicited about the same number of coronavirus-foiling antibodies as the next-generation vaccine based on unmodified mRNA. But importantly, it also caused many fewer side effects, allowing CureVac to safely up the dosage for maximal immune protection. The data in favor of modifications were “very compelling”, says Phil Dormitzer, global head of vaccines research and development at GSK. “Modified RNA has really proven itself.” CureVac, in collaboration with GSK, is now moving forward with high-dose modified mRNA for both of its COVID-19 and flu vaccine programs.

Could computer models be the key to better COVID vaccines?

CureVac is not the first firm to run trials of mRNA vaccine candidates, with and without modifications. Other companies, including BioNTech in Mainz, Germany, and Sanofi, headquartered in Paris, have also done so — but none had disclosed their full results. “That’s why the CureVac study is so helpful,” says vaccinologist Norbert Pardi at the University of Pennsylvania Perelman School of Medicine in Philadelphia.

A previously discovered "hot Jupiter" appears to be spiraling into its host star, new data shows.

Kepler-1658 b, a hot Jupiter with a storied history, appears to be spiraling toward its host star – and will likely crash to the star’s surface in less than 3 million years.

Key facts: The deep-fried planet, about as big as our own Jupiter but nearly six times its mass, has a long history with scientists and the public. It was the first planet candidate to be identified by NASA’s Kepler Space Telescope in 2009 – the same year it launched. Those were the early days of surveying exoplanets, or planets around other stars. But doubts about the detection cast Kepler-1658 b into the dustbin of “false positive” exoplanet discoveries. New software and analytical techniques rescued the planet in the years that followed, and placed it in the “confirmed” column by 2019.

Now, only a few short years after it was brought back into the fold, Kepler-1658 b has been consigned to probable oblivion by the latest observations of its orbital path.

Details: Exoplanet scientists deploy an array of techniques to study distant planets, from the “transit” method Kepler used to discover Kepler-1658 b – measuring the tiny dip in starlight as the planet crosses in front of its star – to carefully tracking the wobbles of parent stars, revealing the gravitational tugs of unseen planets. And these techniques continue to grow in refinement. By measuring subtle shifts in the timing of Kepler 1658 b’s transits across the face of its swollen star, a study team determined the planet was arriving earlier than its expected transit times – a key piece of evidence that the planet’s orbit is shrinking.

The team also showed that the giant planet’s decaying orbit is best explained by tidal forces from its parent star, which, as it moves toward its end stage, is expanding. That expansion increases the gravitational interaction between the star and the planet; the star, in a sense, robs the planet’s orbit of its energy. Our home planet also could meet its end when the Sun grows so large it might swallow Earth, billions of years from now.

Kepler-1658 b also appears to be brighter and hotter than scientists expected, another possible result of tidal interaction with the star.

A newly discovered comet could be visible to the naked eye as it shoots past Earth and the Sun in the coming weeks for the first time in 50,000 years, astronomers have said.

The comet is called C/2022 E3 (ZTF) after the Zwicky Transient Facility, which first spotted it passing Jupiter in March last year. After traveling from the icy reaches of our Solar System it will come closest to the Sun on January 12, 2023 and pass nearest to Earth on February 1, 2023. It will be easy to spot with a good pair of binoculars and likely even with the naked eye, provided the sky is not too illuminated by city lights or the Moon.

The comet "will be brightest when it is closest to the Earth", Thomas Prince, a physics professor at the California Institute of Technology who works at the Zwicky Transient Facility, told AFP.

Made of ice and dust and emitting a greenish aura, the comet is estimated to have a diameter of around a kilometer, said Nicolas Biver, an astrophysicist at the Paris Observatory. That makes it significantly smaller than NEOWISE, the last comet visible with an unaided eye, which passed Earth in March 2020, and Hale–Bopp, which swept by in 1997 with a potentially life-ending diameter of around 60 kilometers.

But the newest visit will come closer to Earth, which "may make up for the fact that it is not very big", Biver said. While the comet will be brightest as it passes Earth in early February, a fuller moon could make spotting it difficult. For the Northern Hemisphere, Biver suggested the last week of January, when the comet passes between the Ursa Minor and Ursa Major constellations. The new moon during the weekend of January 21-22 offers a good chance for stargazers, he said. "We could also get a nice surprise and the object could be twice as bright as expected," Biver added.